Comparison content gets recommended when it helps a buyer choose between real options, with tests and measurements the reader can check. A comparison that names the products, states the criteria up front, shows how each one performed, and gives a verdict per use case is useful. A comparison that lists specs copied from manufacturer pages and declares a winner with no testing is not, and an AI system has no reason to prefer it over the next page that did the work.
There is no vendor formula for what AI recommends. No search engine or assistant publishes a scoring formula for recommendations, and no markup or file makes a recommendation happen. Google's guide to generative AI features says its systems retrieve relevant pages from the Search index and generate a response grounded in what they retrieved, with links to the pages that support the answer. Your comparison page has to be the page worth retrieving.
Pick a comparison a buyer actually makes
Start with a choice your buyers already face. The strongest comparisons are narrow: two or three products, one use case, one buyer. "Espresso machine A versus espresso machine B for a small kitchen" is a real decision. "The best coffee makers" is not a decision, it is a shelf, and a page that tries to rank every machine on the market usually ends up saying little about any of them.
Name the buyer in the first paragraph. A machine that suits a beginner who wants one button is not the machine that suits a hobbyist who wants to control pressure, and a comparison that does not say who it is for cannot give a useful verdict. State the use case, the budget range, and the constraint that matters, such as counter space, noise, or maintenance.
Keep the set small. Two to five products is a workable range. More than that, and the page becomes a catalog with a winner badge, not a comparison. If you sell twenty machines, compare the three your buyers ask about most, and link to the rest.
Test them yourself, and show the work
First-hand testing is what separates a comparison worth citing from a spec sheet with opinions. Google's guide to helpful content says that with product reviews, it builds trust when readers understand how many products were tested, what the test results were, and how the tests were conducted, with evidence of the work involved, such as photographs. The same test applies to comparison content.
For each product, record the same facts, measured the same way:
- The setup. How long it took, what was in the box, and what was missing.
- The result that matters. Shot time, temperature, noise, or whatever the buyer will notice first, measured with the same method for each product.
- The daily use. Cleaning, refilling, descaling, and the part that wears first.
- The cost over time. The price, plus the filters, pods, or parts the buyer will keep buying.
- The limits. What each product cannot do, stated plainly.
Photograph the test. A photo of the machine on your counter, the shot you pulled, or the part that failed is evidence a reader can see and a model can treat as first-hand. Do not use the manufacturer's product photos as your only images and call the page a review.
If you did not test a product, say so. A comparison that includes a product you never handled should label it clearly, give the source of the facts you state about it, and keep it out of the verdict. Presenting a spec-sheet read as a hands-on test is a factual error, and Google's helpful content guide asks whether content has easily verified factual errors.
State the criteria before the verdict
Put the criteria at the top of the page, before any product appears. A buyer should know what you measured and why it matters before reading the results. Criteria for an espresso machine might be shot quality, ease of use, cleaning effort, counter footprint, and running cost. Criteria for a mattress might be firmness, motion transfer, edge support, and return terms. The criteria are the page's argument, and a page with no stated criteria has no argument.
Score each product against the same criteria, in the same order, with the same words. If one product gets "easy to clean" and another gets "simple maintenance," a reader cannot compare the two. Use one scale and one vocabulary across the page.
Give a verdict per use case, not one overall winner. "Best for a beginner with a small kitchen" and "best for a hobbyist who wants control" are two different answers, and both can be true. A single winner declared for every buyer is a sign the page was written to sell one product, not to help a buyer choose.
Disclose the commercial facts. If you sell one of the products, say so. If a product was sent to you free, say so. A comparison that hides its commercial interest is not trustworthy, and a buyer who discovers the interest later will not trust the verdict.
Keep every fact current and checkable
A comparison is only as good as its prices and its specs. Check the price, the availability, and the key specifications for every product on the page, and update them when they change. A comparison that recommends a product at a price it no longer sells for sends a buyer to your site with the wrong expectation.
Link each product to its own product page, and link claims to their source where the source is not your own test. If a manufacturer states a wattage or a capacity, link to the manufacturer page that states it. If a review site measured something you did not, name the site and link to it. Google's helpful content guide asks whether content that draws on other sources avoids simply copying them and instead provides substantial additional value and originality. Your added value is the test, not the copied spec.
Do not copy the comparison from another site and change the product names. Google's ranking systems include original content systems that work to show original content prominently ahead of pages that merely cite it. A copied comparison is the page that gets passed over.
What to leave out
Leave out a winner chosen before the test. If the verdict was decided by margin, by stock, or by a sponsorship, the page is an ad, not a comparison, and a reader will treat it as one.
Leave out products you cannot describe honestly. A product with no price, no availability, and no test result does not belong in the comparison. It belongs on a product page of its own.
Leave out vague verdicts. "Great for everyone" and "a solid choice" say nothing a buyer can act on. Replace each one with the use case and the reason: who it suits, and which measurement supports that.
Leave out thin comparison pages built to capture a search query. Google's guide to generative AI features warns against creating separate pages for every possible variation of a question mainly to manipulate rankings or generative responses, and says a high quantity of pages does not make a site higher quality. One tested comparison per real decision beats ten untested ones.
The fundamentals
These mechanics are shared across every engine and covered in full in their own guides:
- •The llms.txt format and a complete sample, see our guide to writing a good llms.txt file
- •The markup this engine reads, see our product schema guide
- •The complete crawler rule reference, see our robots.txt guide for AI search
Say you run a home espresso shop
Say you sell home espresso machines, and your comparison page lists six machines with specs copied from manufacturer pages and a "best overall" badge on the machine with the highest margin. A buyer asks an assistant which espresso machine suits a beginner with a small kitchen. The assistant retrieves your page, but the page never says who each machine suits, never states how any of them were tested, and never gives a cleaning or running cost, so it quotes a smaller shop whose page tested three machines side by side, timed the shots, photographed the cleanup, and gave a verdict per use case. You rebuild the page around those three machines, with your own test notes, your own photos, the criteria up front, and a verdict for the beginner and a verdict for the hobbyist. Nothing else on the site changes. The next time the question is asked, your page is the one with evidence behind the answer.
What this guide does not cover
This guide covers comparison content you write yourself. It does not cover the Product snippet markup for editorial review pages, which is our guide to adding product schema, and it does not cover getting the page indexed, getting crawlers to it, or measuring whether it appears in AI answers, which is monitoring work. A tested comparison makes a page recommendable. It does not promise a recommendation.
Check your own brand now
Run the free scan. It takes seconds, no signup, and shows whether AI crawlers can read your comparison pages today: aizii.ai/aeo-scan
Run the free scanSources
- Google Search Central, Optimizing your website for generative AI features on Google Search (specific): https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Google Search Central, Creating helpful, reliable, people-first content (specific): https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- Google Search Central, A guide to Google Search ranking systems (specific): https://developers.google.com/search/docs/appearance/ranking-systems-guide
- Google Search Central, Product snippet structured data (specific): https://developers.google.com/search/docs/appearance/structured-data/product-snippet
- Google Search Central, Product structured data (specific): https://developers.google.com/search/docs/appearance/structured-data/product
- Schema.org, Review type (specific): https://schema.org/Review
Note: links marked specific point to the page that supports the claim in this guide. Links marked general docs point to the vendor's general docs section for that topic, not to a page that proves every step. Check the vendor docs for the current path before relying on a menu name or click path in this guide.
